AI shopping needs new protocols
An Ecommerce Fastlane explainer argues that AI shopping will rely on infrastructure standards—ACP, UCP and MCP—so retailers can be discovered and transacted through assistants, and that OpenAI’s Instant Checkout experiment exposed gaps. The piece says most retailers still don’t support these protocols, meaning discovery and checkout success for AI assistants will depend on backend standards, not just front‑end design. That shifts the product problem toward instrumentation, eligibility, and checkout reliability for AI-driven commerce. (ecommercefastlane.com)
OpenAI and Google are both trying to turn chatbots into shopping carts, but the bottleneck is not the chatbot window people see. The bottleneck is whether a retailer’s catalog, checkout, and inventory systems speak the machine-readable standards those assistants can use. (developers.openai.com, developers.google.com) OpenAI’s standard is called the Agentic Commerce Protocol, which is an open specification built with Stripe for purchases inside tools like ChatGPT. Stripe says it lets a seller expose checkout to compatible apps, while OpenAI says it lets ChatGPT ingest structured catalog data and understand merchant inventory. (docs.stripe.com, developers.openai.com) Google’s standard is called the Universal Commerce Protocol, and Google describes it as an open standard for “AI Mode” and Gemini that starts with direct buying. Google’s developer guide says the goal is to turn AI interactions into sales without custom one-off integrations for each retailer. (developers.google.com, developers.googleblog.com) A third piece sits underneath both shopping flows. The Model Context Protocol is the plumbing standard for connecting an artificial intelligence assistant to outside tools and data, and Anthropic donated it to the Linux Foundation’s Agentic AI Foundation in December 2025 with support from OpenAI, Google, Microsoft, Amazon Web Services, Cloudflare, and Bloomberg. (anthropic.com, aaif.io) That means these three standards do different jobs. The Model Context Protocol is the adapter that lets an assistant reach a store’s systems, the Universal Commerce Protocol is Google’s shopping language across discovery and purchase, and the Agentic Commerce Protocol is the checkout language OpenAI and Stripe built for in-chat buying. (developers.google.com, docs.stripe.com, anthropic.com) The reason retailers suddenly care is that old web shopping was built for humans clicking pages, while assistant shopping is built for software calling endpoints. A product page with pretty photos can still fail if the assistant cannot verify stock, resolve a size variant, link a loyalty account, or hand off a payment token cleanly. (docs.stripe.com, blog.google) Google’s recent Universal Commerce Protocol updates show what the real work looks like. In March 2026, Google said the protocol added multi-item cart saving, real-time price and inventory access, identity linking for loyalty benefits, and simpler onboarding through Merchant Center. (blog.google) OpenAI’s side shows the same pattern from the checkout end. Stripe’s launch post for Instant Checkout in ChatGPT said the feature ran on the Agentic Commerce Protocol, and Stripe’s current docs still label its broader agentic commerce tooling and protocol integration as private preview, which tells you the rails are still being built in public. (stripe.com, docs.stripe.com, docs.stripe.com) The Ecommerce Fastlane piece argues that OpenAI’s Instant Checkout experiment exposed a hard truth: discovery and conversion break when merchant backends are not instrumented for assistants. Its practical advice starts with unglamorous checks like platform admin access, Google Merchant Center access, robots.txt review, and auditing top product listings. (ecommercefastlane.com) So the new product problem is not “how do we redesign the storefront for artificial intelligence.” It is “can an assistant reliably find the right item, confirm eligibility, carry identity, and complete checkout without hitting a dead end in the merchant stack.” (ecommercefastlane.com, blog.google, docs.stripe.com) The retailers that win this shift may look boring from the outside. They will have clean catalog data, stable inventory feeds, machine-readable checkout flows, and protocol support that makes them easy for ChatGPT, Gemini, and future assistants to trust and transact with. (developers.openai.com, developers.google.com, ecommercefastlane.com)
